Subnet Weights
Subnet weights describe how Bittensor compares subnets for cross-subnet emission allocation. They are tied to net TAO flows and the flow-based emission model (Glossary: Subnet Weights, Emission: Distribution Across Subnets).
The term is about relative emission share across subnets. It is not the same as validator weights, weight vectors, or weights used inside a single subnet’s consensus process.
Flow Signal
Subnet weights are tied to net TAO flow. Net flow is used to guide cross-subnet allocation, making subnet weights an allocation signal rather than a direct measure of work quality inside one subnet (Emission: Distribution Across Subnets, Glossary: Subnet Weights).
This keeps the term tied to cross-subnet movement. A subnet can have miners, validators, and its own incentive mechanism, but subnet weights compare subnet-level flow signals.
Smoothing Context
The flow signal is smoothed with an exponential moving average. Smoothing keeps a brief movement from being treated as a durable allocation signal (Exponential Moving Averages, Emission: Distribution Across Subnets).
That distinction helps keep subnet weights from being read as a direct event log. They are part of the emission-allocation model, not a complete record of every individual flow event.
Emission Allocation
Subnet weights belong to the cross-subnet allocation stage. They help answer how much relative emission importance is assigned to a subnet before rewards are resolved inside that subnet (Emission: Injection and Distribution, Yuma Consensus).
This keeps allocation and distribution layers separate. Subnet weights help determine relative importance across subnets, while within-subnet consensus determines how rewards are resolved among miners and validators.
Validator Weight Contrast
Subnet weights and validator weights use similar vocabulary but refer to different scopes. Subnet weights compare subnets; validator weights compare miner outputs inside one subnet’s consensus context (Glossary: Validator Weights, Glossary: Subnet Weights).
Reading the object being weighted usually resolves the ambiguity. Subnet weights are subnet-level; validator weights are miner-evaluation signals inside a subnet.
Alpha Token Context
Subnet weights are related to alpha-token tokenomics, but they are not alpha tokens themselves. Alpha tokens are subnet-specific assets, while subnet weights describe relative cross-subnet allocation (Glossary: Alpha Tokens, Emission: Distribution Across Subnets).
That distinction is useful when reading Dynamic TAO material. Alpha tokens name subnet assets; subnet weights name relative allocation across subnets.
Relationship to Yuma Consensus
Subnet Weights and Yuma Consensus describe related parts of Bittensor’s incentive system. Yuma Consensus is the on-chain process that aggregates validator weight signals within a subnet into miner incentives and validator dividends, applying consensus clipping, bonding, and emission calculation (Yuma Consensus).
For readers, subnet Weights names a specific part of that incentive picture, while Yuma Consensus names the consensus process that turns validator weights into the resulting incentives and dividends.
Reader Boundary
Use subnet weights when the focus is cross-subnet emission share. Use validator weights, weight vectors, or weight matrices when the focus is narrower within-subnet consensus signals (Glossary: Subnet Weights, Glossary: Validator Weights).
The term sits between Dynamic TAO’s broad subnet tokenomics and the more specific weight terms used inside consensus.
Development Stage Context
The Introduction to Bittensor describes subnet development as moving from localnet to testnet and then mainnet. The subnet weights concept applies across the Bittensor lifecycle: flow-based allocation signals are relevant on testnet for non-production evaluation and on mainnet for live cross-subnet emission distribution.
The Bittensor Networks reference separates mainnet, testnet, and localnet. Subnet weight examples or emission allocation outcomes from one environment should not be read as representing production cross-subnet allocation in another environment.
Consensus Scope
Subnet weights should stay attached to the subnet and consensus context where they are used. Yuma Consensus explains validator weighting, while the metagraph source keeps subnet state tied to a selected netuid (Yuma Consensus, Subnet Metagraph).
For readers, a subnet weight claim is incomplete without the subnet context. The same vocabulary can describe different participants or mechanisms depending on the selected netuid and network state.
Negative Net Flow Can Zero Out Injection
Emission: Distribution across subnets documentation describes the flow-based model as rewarding subnets with positive net TAO inflows from staking activity and assigning zero emissions to subnets whose offset net flow falls to zero after clipping. Subnet weights in that reading are relative importance signals inside a competitive cross-subnet allocation step, not guaranteed emission regardless of staking behavior.
That clipping rule keeps subnet weights tied to sustained staking direction. A subnet can lose injection share when outflows dominate inflows long enough for its smoothed net flow to fall below the model’s lower bound.
Registration Burns Do Not Count as User Inflow
The same emission reference notes an exception to user-flow tracking:
registration burns from
burned_register are excluded from net inflows, while stake and unstake operations are counted.
Subnet weights therefore respond to staking movement rather than to UID purchase burns alone.
That boundary prevents registration spending from being read as the same signal as ongoing stake support when interpreting cross-subnet allocation.
Cross-Subnet Share Precedes Within-Subnet Yuma
Official emission documentation separates injection from distribution. Injection adds TAO to subnet pools each block using flow-based shares, while distribution at each tempo boundary allocates accumulated rewards inside a subnet through Yuma Consensus (Emission: Injection and Distribution).
Subnet weights belong to the first stage. They compare subnets for relative TAO injection share before validator weights, rank, and dividends are resolved inside any one subnet’s consensus pass.